The rapid adoption of generative artificial intelligence has transformed chatbots such as ChatGPT, Google Gemini, Microsoft Copilot, and Meta AI into everyday tools. Millions of people now rely on them for advice about emotions, relationships, and mental well-being because they are available around the clock, easy to access, anonymous, and nonjudgmental.
However, a new analysis published in the Journal of Psychopathology and Clinical Science suggests that unsupervised use of general-purpose AI chatbots for mental health support may unintentionally reinforce harmful psychological patterns rather than reduce them.
AI chatbots and mental healthLarge language model (LLM)-based chatbots are increasingly being used for personal conversations rather than work-related tasks. Studies indicate that around 70% of chatbot interactions involve everyday issues, including emotional concerns, relationships, and mental health.
Research also suggests that nearly half of adults diagnosed with a mental health condition have sought support from AI chatbots, often without informing their healthcare providers. Many users report preferring AI because it is constantly available, anonymous, and does not judge them.
While some AI applications have been specifically designed as mental health tools and have shown benefits in clinical studies, general-purpose chatbots such as ChatGPT and Gemini were not developed to provide psychiatric care and do not include the specialized crisis protocols or clinical safeguards used in professional mental health services.
Five potential risks identified by researchersAfter reviewing the available scientific evidence, the author identified five ways in which unsupervised chatbot use could unintentionally contribute to ongoing mental health problems.
1. Delaying professional treatment
People may begin relying on AI as an inexpensive and convenient alternative to qualified mental health professionals. Although chatbots may provide temporary emotional support, they could create a false sense of receiving adequate care and delay appropriate diagnosis or treatment.
2. Reinforcing compulsive reassurance-seeking
Because AI chatbots are generally designed to be helpful and agreeable, they may repeatedly reassure users experiencing anxiety or obsessive-compulsive disorder (OCD). Instead of reducing anxiety, this repeated reassurance may strengthen compulsive reassurance-seeking behaviors.
3. Increasing social isolation
For individuals who struggle with loneliness or social anxiety, conversations with AI may feel safer than interactions with other people because chatbots do not criticize, reject, or misunderstand users. Over time, this could encourage people to substitute AI interactions for real-life relationships.
4. Strengthening distorted beliefs
Researchers also warn that AI systems may unintentionally reinforce inaccurate or even delusional beliefs if they consistently agree with users rather than challenge irrational thinking. The author refers to this potential phenomenon as "AI psychosis," although additional research is needed to better understand this risk.
5. Reducing confidence in personal decision-making
Frequent dependence on AI for everyday choices may gradually weaken users' trust in their own judgment, making them increasingly reliant on chatbot advice instead of developing independent decision-making skills.
More evidence is neededThe author emphasizes that current evidence remains limited. Many published findings are based on observational studies, anecdotal reports, or individual case studies rather than large clinical trials.
Further research is needed to determine how AI affects mental health over the long term, particularly among vulnerable individuals. The paper also calls for stronger safety measures, closer clinical oversight, and additional safeguards from AI developers to minimize potential risks while preserving the benefits these technologies can provide.
Follow NEWS.am Medicine on Facebook and Twitter